gigaparsec/gigaparsec.go

package gigaparsec

import (
	"errors"
	"fmt"
	"io"
	"slices"

	"git.codemonkeysoftware.net/b/gigaparsec/cursor"
)

type Result[In, Out any] struct {
	consumed, failed bool
	value            Out
	next             State[In]
	message          Message
}

func Fail[In, Out any](consumed bool, msg Message) Result[In, Out] {
	return Result[In, Out]{
		consumed: consumed,
		failed:   true,
		message:  msg,
	}
}

func (r Result[In, Out]) Failed() (ok, consumed bool, msg Message) {
	ok = r.failed
	if ok {
		consumed = r.consumed
		msg = r.message
	}
	return
}

func Succeed[In, Out any](consumed bool, value Out, next State[In], msg Message) Result[In, Out] {
	return Result[In, Out]{
		failed:   false,
		value:    value,
		consumed: consumed,
		next:     next,
		message:  msg,
	}
}

func (r Result[In, Out]) Succeeded() (ok, consumed bool, value Out, next State[In], msg Message) {
	ok = !r.failed
	if ok {
		consumed = r.consumed
		value = r.value
		next = r.next
		msg = r.message
	}
	return
}

func (r Result[In, Out]) Consumed() bool {
	return r.consumed
}

func (r Result[In, Out]) Consume(consumed bool) Result[In, Out] {
	r.consumed = consumed
	return r
}

type Message struct {
	Pos      uint64
	Got      string
	Expected []string
}

func (m Message) expect(s string) Message {
	m.Expected = []string{s}
	return m
}

func MessageOK(pos uint64) Message { return Message{Pos: pos} }

func MessageEnd(pos uint64) Message { return Message{Pos: pos, Got: "end of input"} }

func MakeState[In any](c cursor.Cursor[In]) State[In] {
	return State[In]{cursor: c}
}

type State[In any] struct {
	cursor cursor.Cursor[In]
}

func (s State[In]) Cursor() cursor.Cursor[In] {
	return s.cursor
}

func (s State[In]) Read(dst []In) (n uint64, next State[In], err error) {
	n, c, err := s.cursor.Read(dst)
	return n, State[In]{cursor: c}, err
}

func (s State[In]) Pos() uint64 {
	return s.cursor.Pos()
}

func (s State[In]) At(pos uint64) State[In] {
	return State[In]{cursor: s.cursor.At(pos)}
}

type Parser[In, Out any] func(State[In]) (Result[In, Out], error)

func Return[In, Out any](value Out) Parser[In, Out] {
	return func(state State[In]) (Result[In, Out], error) {
		return Succeed(false, value, state, MessageOK(state.Pos())), nil
	}
}

func Satisfy[T any](pred func(T) bool) Parser[T, T] {
	return func(state State[T]) (Result[T, T], error) {
		token := make([]T, 1)
		n, next, err := state.Read(token)
		if errors.Is(err, io.EOF) {
			return Fail[T, T](false, MessageEnd(state.Pos())), nil
		}
		if err != nil {
			return Result[T, T]{}, err
		}
		if n != 1 {
			panic(fmt.Sprintf("expected 1 element from Read, but got %d", n))
		}
		if pred(token[0]) {
			return Succeed(true, token[0], next, MessageOK(state.Pos())), nil
		}
		return Fail[T, T](false, Message{
			Pos: state.Pos(),
			Got: fmt.Sprint(token),
		}), nil
	}
}

func Slice[T comparable](s []T) Parser[T, []T] {
	expected := fmt.Sprint(s)
	return func(state State[T]) (reply Result[T, []T], err error) {
		token := make([]T, len(s))
		_, next, err := state.Read(token)
		if errors.Is(err, io.EOF) {
			return Fail[T, []T](false, MessageEnd(state.Pos())), nil
		}
		if err != nil {
			return Result[T, []T]{}, err
		}
		if !slices.Equal(s, token) {
			return Fail[T, []T](false, Message{
				Pos:      state.Pos(),
				Got:      fmt.Sprint(token),
				Expected: []string{expected},
			}), nil
		}
		return Succeed(true, token, next, MessageOK(state.Pos())), nil
	}
}

func Bind[In, A, B any](p Parser[In, A], f func(A) Parser[In, B]) Parser[In, B] {
	return func(input State[In]) (Result[In, B], error) {
		resultA, err := p(input)
		if err != nil {
			return Result[In, B]{}, err
		}
		if ok, consumed, msg := resultA.Failed(); ok {
			return Fail[In, B](consumed, msg), nil
		}

		_, consumedA, valueA, next, _ := resultA.Succeeded()
		resultB, err := f(valueA)(next)
		if err != nil {
			return Result[In, B]{}, err
		}
		return resultB.Consume(consumedA || resultB.Consumed()), nil
	}
}

func Choose[In, Out any](p Parser[In, Out], ps ...Parser[In, Out]) Parser[In, Out] {
	// TODO Check this against the Parsec paper again, and simplify it.
	all := append([]Parser[In, Out]{p}, ps...)
	return func(input State[In]) (Result[In, Out], error) {
		expecteds := make([][]string, 0, len(all))
		var value Out
		var got string
		var failed bool
		for _, q := range all {
			result, err := q(input)
			if err != nil {
				return Result[In, Out]{}, err
			}
			if result.Consumed() {
				return result, nil
			}
			var qMsg Message
			if isFailure, _, msg := result.Failed(); isFailure {
				qMsg = msg
				failed = true
			} else {
				_, _, qValue, _, msg := result.Succeeded()
				if failed {
					value = qValue
					failed = false
				}
				qMsg = msg
			}
			if got == "" {
				got = qMsg.Got
			}
		}
		msg := Message{Pos: input.Pos(), Got: got, Expected: slices.Concat(expecteds...)}
		if failed {
			return Fail[In, Out](false, msg), nil
		}
		return Succeed(false, value, input, msg), nil
	}
}

// Try behaves identically to p, except that if p returns an error,
// Try will pretend that no input was consumed. This allows infinite
// lookahead: Since Choose only calls another parser when the previous
// parser consumed nothing, Try will allow backing out of a complex
// parser that partially succeeded.
func Try[In, Out any](p Parser[In, Out]) Parser[In, Out] {
	return func(input State[In]) (Result[In, Out], error) {
		result, err := p(input)
		if err != nil {
			return result, err
		}
		if failed, _, msg := result.Failed(); failed {
			return Fail[In, Out](false, msg), nil
		}
		return result, nil
	}
}

func Label[In, Out any](p Parser[In, Out], l string) Parser[In, Out] {
	return func(input State[In]) (reply Result[In, Out], err error) {
		result, err := p(input)
		if err != nil || result.Consumed() {
			return result, err
		}
		if succeeded, _, value, next, msg := result.Succeeded(); succeeded {
			return Succeed(false, value, next, msg.expect(l)), nil
		}
		_, _, msg := result.Failed()
		return Fail[In, Out](false, msg.expect(l)), nil
	}
}

func Map[In, Out1, Out2 any](p Parser[In, Out1], f func(Out1) Out2) Parser[In, Out2] {
	return Bind(p, func(out Out1) Parser[In, Out2] {
		return Return[In](f(out))
	})
}

func End[In any](s State[In]) (reply Result[In, struct{}], err error) {
	_, _, err = s.cursor.Read([]In{})
	if errors.Is(err, io.EOF) {
		return Succeed(true, struct{}{}, s, MessageOK(s.Pos())), nil
	}
	if err != nil {
		return Result[In, struct{}]{}, fmt.Errorf("End: unexpected error: %w", err)
	}
	return Fail[In, struct{}](false, Message{
		Pos:      s.Pos(),
		Got:      "",
		Expected: []string{"end of input"},
	}), nil
}